Start work on multipart form parser

src/http/middleware.zig

@@ -587,13 +587,15 @@ const BaseContentType = enum {
     json,
     url_encoded,
     octet_stream,
+    multipart_formdata,
 
     other,
 };
 
 fn parseBodyFromRequest(comptime T: type, content_type: ?[]const u8, reader: anytype, alloc: std.mem.Allocator) !T {
     // Use json by default for now for testing purposes
-    const parser_type = matchContentType(content_type) orelse .json;
+    const eff_type = content_type orelse "application/json";
+    const parser_type = matchContentType(eff_type);
     const buf = try reader.readAllAlloc(alloc, 1 << 16);
     defer alloc.free(buf);
 
@@ -608,21 +610,32 @@ fn parseBodyFromRequest(comptime T: type, content_type: ?[]const u8, reader: any
             error.NoQuery => error.NoBody,
             else => err,
         },
+        .multipart_formdata => {
+            const param_string = std.mem.split(u8, eff_type, ";").rest();
+            const params = query_utils.parseQuery(alloc, struct {
+                boundary: []const u8,
+            }, param_string) catch |err| return switch (err) {
+                error.NoQuery => error.MissingBoundary,
+                else => err,
+            };
+            defer util.deepFree(alloc, params);
+
+            try @import("./multipart.zig").parseFormData(params.boundary, reader, alloc);
+        },
         else => return error.UnsupportedMediaType,
     }
 }
 
 // figure out what base parser to use
-fn matchContentType(hdr: ?[]const u8) ?BaseContentType {
-    if (hdr) |h| {
-        if (std.ascii.eqlIgnoreCase(h, "application/x-www-form-urlencoded")) return .url_encoded;
-        if (std.ascii.eqlIgnoreCase(h, "application/json")) return .json;
-        if (std.ascii.eqlIgnoreCase(h, "application/octet-stream")) return .octet_stream;
+fn matchContentType(hdr: []const u8) BaseContentType {
+    const trimmed = std.mem.sliceTo(hdr, ';');
+    if (std.ascii.eqlIgnoreCase(trimmed, "application/x-www-form-urlencoded")) return .url_encoded;
+    if (std.ascii.eqlIgnoreCase(trimmed, "application/json")) return .json;
+    if (std.ascii.endsWithIgnoreCase(trimmed, "+json")) return .json;
+    if (std.ascii.eqlIgnoreCase(trimmed, "application/octet-stream")) return .octet_stream;
+    if (std.ascii.eqlIgnoreCase(trimmed, "multipart/form-data")) return .multipart_formdata;
 
-        return .other;
-    }
-
-    return null;
+    return .other;
 }
 
 /// Parses a set of body arguments from the request body based on the request's Content-Type
@@ -674,6 +687,10 @@ test "parseBodyFromRequest" {
     };
     try testCase("application/json", "{\"id\": 3}", Struct{ .id = 3 });
     try testCase("application/x-www-form-urlencoded", "id=3", Struct{ .id = 3 });
+
+    try testCase("multipart/form-data; ",
+        \\
+    , Struct{ .id = 3 });
 }
 
 test "parseBody" {

src/http/multipart.zig

@@ -0,0 +1,154 @@
+const std = @import("std");
+
+const max_boundary = 70;
+
+const FormField = struct {
+    //name: []const u8,
+    //disposition: []const u8,
+    //filename: ?[]const u8,
+    //charset: ?[]const u8,
+    value: []const u8,
+};
+
+const FormFieldResult = struct {
+    field: FormField,
+    more: bool,
+};
+
+fn isFinalPart(peek_stream: anytype) !bool {
+    const reader = peek_stream.reader();
+    var buf: [2]u8 = undefined;
+    const end = try reader.readAll(&buf);
+    const end_line = buf[0..end];
+    const terminal = std.mem.eql(u8, end_line, "--");
+    if (!terminal) try peek_stream.putBack(end_line);
+
+    // Skip whitespace
+    while (true) {
+        const b = reader.readByte() catch |err| switch (err) {
+            error.EndOfStream => {
+                if (terminal) break else return error.InvalidMultipartBoundary;
+            },
+            else => return err,
+        };
+
+        if (std.mem.indexOfScalar(u8, " \r\n", b) == null) {
+            try peek_stream.putBackByte(b);
+            break;
+        }
+    }
+
+    return terminal;
+}
+
+fn parseFormField(boundary: []const u8, peek_stream: anytype, alloc: std.mem.Allocator) !FormFieldResult {
+    const reader = peek_stream.reader();
+
+    // TODO: refactor
+    var headers = try @import("./request/parser.zig").parseHeaders(alloc, reader);
+    defer headers.deinit();
+
+    std.debug.print("disposition: {?s}\n", .{headers.get("Content-Disposition")});
+
+    var value = std.ArrayList(u8).init(alloc);
+    errdefer value.deinit();
+
+    line_loop: while (true) {
+        // parse crlf--
+        var buf: [4]u8 = undefined;
+        try reader.readNoEof(&buf);
+        if (!std.mem.eql(u8, &buf, "\r\n--")) {
+            try value.append(buf[0]);
+            try peek_stream.putBack(buf[1..]);
+            var ch = try reader.readByte();
+            while (ch != '\r') : (ch = try reader.readByte()) try value.append(ch);
+
+            try peek_stream.putBackByte(ch);
+            continue;
+        }
+
+        for (boundary) |ch, i| {
+            const b = try reader.readByte();
+
+            if (b != ch) {
+                try value.appendSlice("\r\n--");
+                try value.appendSlice(boundary[0 .. i + 1]);
+                continue :line_loop;
+            }
+        }
+
+        // Boundary parsed. See if its a terminal or not
+        break;
+    }
+
+    const terminal = try isFinalPart(peek_stream);
+
+    return FormFieldResult{
+        .field = .{ .value = value.toOwnedSlice() },
+        .more = !terminal,
+    };
+}
+
+pub fn parseFormData(boundary: []const u8, reader: anytype, alloc: std.mem.Allocator) !void {
+    if (boundary.len > max_boundary) return error.BoundaryTooLarge;
+
+    var stream = std.io.peekStream(72, reader);
+    {
+        var buf: [72]u8 = undefined;
+        const count = try stream.reader().readAll(buf[0 .. boundary.len + 2]);
+        var line = buf[0..count];
+        if (line.len != boundary.len + 2) return error.InvalidMultipartBoundary;
+        if (!std.mem.startsWith(u8, line, "--")) return error.InvalidMultipartBoundary;
+        if (!std.mem.endsWith(u8, line, boundary)) return error.InvalidMultipartBoundary;
+
+        if (try isFinalPart(&stream)) return;
+    }
+
+    while (true) {
+        const field = try parseFormField(boundary, &stream, alloc);
+        alloc.free(field.field.value);
+
+        if (!field.more) return;
+    }
+}
+fn toCrlf(comptime str: []const u8) []const u8 {
+    comptime {
+        var buf: [str.len * 2]u8 = undefined;
+
+        @setEvalBranchQuota(@intCast(u32, str.len * 2)); // TODO: why does this need to be *2
+
+        var buf_len: usize = 0;
+        for (str) |ch| {
+            if (ch == '\n') {
+                buf[buf_len] = '\r';
+                buf_len += 1;
+            }
+
+            buf[buf_len] = ch;
+            buf_len += 1;
+        }
+
+        return buf[0..buf_len];
+    }
+}
+
+test "parseFormData" {
+    const body = toCrlf(
+        \\--abcd
+        \\Content-Disposition: form-data; name=first
+        \\
+        \\content
+        \\--abcd
+        \\content-Disposition: form-data; name=second
+        \\
+        \\no content
+        \\--abcd
+        \\content-disposition: form-data; name=third
+        \\
+        \\
+        \\--abcd--
+        \\
+    );
+    var stream = std.io.StreamSource{ .const_buffer = std.io.fixedBufferStream(body) };
+    try parseFormData("abcd", stream.reader(), std.testing.allocator);
+}

src/http/request/parser.zig

@@ -93,7 +93,7 @@ fn parseProto(reader: anytype) !http.Protocol {
     };
 }
 
-fn parseHeaders(allocator: std.mem.Allocator, reader: anytype) !Fields {
+pub fn parseHeaders(allocator: std.mem.Allocator, reader: anytype) !Fields {
     var headers = Fields.init(allocator);
 
     var buf: [4096]u8 = undefined;